The Process of Identifying Automobile Joint Failures during the Operation Phase: Data Analytics Based on Association Rules

Abstract

The increasing complexity of vehicle design, the use of new engine types and fuels, and the increasing intelligence of automobiles are making it increasingly difficult to ensure trouble-free operation. Finding faulty parts quickly and accurately is becoming increasingly difficult, as the diagnostic process requires analyzing a great amount of information. Therefore, we propose an approach based on association rules, a machine learning technique, to simplify the defect detection process. To facilitate its use in a real repair company environment, we have developed a web service that allows a repairman to simultaneously identify nodes with a high probability of failure. We have described the structure and working principles of the developed web service, as well as the procedure for its application, which resulted in the discovery of several useful non-trivial rules. We have presented several rules resulting from the use of this interactive tool, which allow repairers to detect possible defects in the relevant components, during the diagnostic process, quickly and easily. These rules are also well supported and can be used by procurement departments to make tactical decisions when selecting the most promising suppliers and manufacturers. The methodology developed allows the evaluation of the effectiveness of changes in the design and technology for the manufacture and operation of individual vehicle components, analyzing the change in the composition of parts combinations over time.